HAL Id: hal-01593556https://hal.archives-ouvertes.fr/hal-01593556

Submitted on 26 Sep 2017

HAL is a multi-disciplinary open accessarchive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come fromteaching and research institutions in France orabroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, estdestinée au dépôt et à la diffusion de documentsscientifiques de niveau recherche, publiés ou non,émanant des établissements d’enseignement et derecherche français ou étrangers, des laboratoirespublics ou privés.

Origin of epoxies embrittlement during oxidative ageingEsteve Ernault, Emmanuel Richaud, Bruno Fayolle

To cite this version:Esteve Ernault, Emmanuel Richaud, Bruno Fayolle. Origin of epoxies embrittlement during oxidativeageing. Polymer Testing, Elsevier, 2017, 63, pp.448-454. �10.1016/j.polymertesting.2017.09.004�. �hal-01593556�

Origin of epoxies embrittlement during oxidative ageing

Esteve Ernault, Emmanuel Richaud*, Bruno FayollePIMM UMR 8006, Arts et M�etiers ParisTech, CNAM, CNRS, 151 bd de l'Hopital, Paris, France

Keywords:Epoxy diamineOxidationFailure properties

* Corresponding author.E-mail address: emmanuel.richaud@ensam.eu (E.

http://dx.doi.org/10.1016/j.polymertesting.2017.09.0040142-9418/© 2017 Elsevier Ltd. All rights reserved.

a b s t r a c t

Thermal oxidation of three epoxy resins prepared from flexible or rigid prepolymers and hardeners wasstudied by monitoring epoxy mechanical and physical changes. The physical changes were followed bymass measurements, glass transition temperature using DSC and sub-glass b transition using DMA. It wasput in evidence that embrittlement is not directly associated to Tg or mass loss changes since epoxynetwork based on isophorone diamine (IPDA) hardeners were shown to undergo mainly a chain scissionat the beginning of exposure process whereas epoxy network based on trioxatridecane diamine (TTDA)hardeners exhibits a crosslinking process with a significant mass loss. The only common feature for bothepoxy systems to understand embrittlement is the drop of amplitude of b transition with oxidation.

1. Introduction

Epoxy family cover a large diversity of polymer networksdiffering by the nature of resins and hardeners commerciallyavailable. Depending on resin and hardener used, range of Tg canvary from 60 �C up to 250 �C for instance tris(4-hydroxyphenyl)methane triglycidyl ether (Tactix 742) and diglycidyl ether ofbisphenol A (Tactix 123), with an aromatic diamine hardener: 4,4-diaminodiphenyl sulfone [1,2].

In many applications such as composites for aeronautical com-ponents, a peculiar concern is associated to long term ageing effectsince oxidation affects strongly the epoxy mechanical properties,especially the failure properties. In this context, epoxy lifetime isoften empirically related to mass loss induced by oxidation. How-ever, to predict lifetime with an accurate manner for all epoxynetworks, embrittlement mechanisms associated to oxidation haveto be deeply studied. Since embrittlement process corresponds tofailure properties drop, let us recall some approaches relating theseproperties with epoxy networks characteristics.

Failure properties, especially strain at break or toughness GIC,are governed by the capacity to promote plastic deformation. Evenif the relationships between structure and plastic deformation forepoxy networks in glassy state have been studied since the se-venties [3e8], the molecular origin for plastic deformation mech-anisms in the epoxy networks still is an open question. The varietyof epoxy networks studied in the literature and the difficulties to

Richaud).

investigate plastic deformation zone can explain this lack ofconsensus in literatures. Several approaches have been proposed tounderstand the molecular aspects of plastic deformation in epoxynetworks:

� Plastic deformation is promoted if glass transition is close tomechanical test temperature. In other words, plasticity is asso-ciated to cooperative mobility.

� Plastic deformation is linked to molar mass between crosslinks(MC). In this approach, plastic deformation obeys the relation-ships proposed in rubber elasticity theory where deformation atfailure is associated to the chain extensibility limit. Indeed,several authors have put in evidence that epoxy network frac-ture energy is proportional to MC

1/2 [9]. However, an oppositetrend has been witnessed by Urbaczewsk-Espuche et al. [10].

� Plastic deformation is associated to secondary transition belowTg, which is often called b transition. The b transition in net-works based on DEGBA is attributed to local motions linked tohydroxyether segments [11]. As a result, plastic deformation(cooperative motions) is activated by this local mobility for agiven state stress.

In this paper, we propose to follow epoxy networks modifica-tions during oxidation in terms of glass transition and b transitionand to relate these modifications to embrittlement process. For thispurpose, we have chosen three epoxy networks: two systemshaving a relatively low Tg (DGEBA/TTDA and DGEBU/IPDA, 69 and60 �C respectively) and a system having a high Tg (DGEBA/IPDA,166 �C). Furthermore, it has been witnessed that among these

systems, one system undergoes mainly chain scission (DGEBA/IPDA) whereas the other undergoes mainly crosslinking [12]. Thefinal objective is then to propose a common scenario to explainembrittlement induced by oxidation for epoxy networks.

2. Experimental

2.1. Materials

Two kinds of prepolymers are used.

- The diglycidyl ether of bisphenol A (DER 332, CAS 1675-54-3eref 31185 supplied by Sigma Aldrich) has a degree of poly-merization n ~ 0 and a number average molecular mass equal to340 g/mol.

- The diglycidyl ether of 1,4-butanediol named DGEBU (CAS 2425-79-8eref 220892 by Sigma Aldrich) has an average molecularmass of 202.25 g/mol.

Two kinds of hardeners have been chosen:

- The isophorone diamine here named IPDA (CAS 2855-13-2eref118184 by Sigma Aldrich, M ¼ 170.3 g/mol).

- The 4,7,10-trioxa-1,13-tridecanediamine here named TTDA (CAS4246-51-9eref 369519 by Sigma Aldrich, M ¼ 220.3 g/mol).

The chemical structure of chemicals is given in Fig. 1.These components were mixed in stoichiometric ratio and fully

cured checked by the total disappearance of exothermal signal andby FTIR from the loss of 914 cm�1 peak attributed to epoxide group.Details of films preparation, thermal characterization of the cureand glass transition of final networks obtained can be found else-where [13]. Thin films have been processed in heating press.

Thermal ageing under atmospheric air was performed inventilated ovens (calibrated at ± 3 �C) at 110 �C. Films having athickness lower than 100 mm to avoid diffusion limited oxidationwere periodically removed for analysis.

2.2. Characterization

2.2.1. Oxidative products concentrationsTo assess oxidative products concentration during exposure, FTIR

spectroscopy in transmission mode was performed on free standingfilms using a Frontier spectrophotometer (PerkinElmer) in the 550to 4000 cm�1 wavenumber range by averaging 16 scans with a4 cm�1 resolution. Spectra were interpreted using the Spectrum

Fig. 1. Chemical structure of (a) diglycidyl ether of bisphenol A (DER332), (b) diglycidylether of 1,4-butanediol (DGEBU) (c) isophorone diamine (IPDA) (d) 4,7,10-trioxa-1,13-tridecanediamine (TTDA).

software (PerkinElmer) in order to determine the absorbance valuefrom which the carbonyl and amide concentrations, the mainoxidation products, were calculated according to the methoddefined elsewhere [13].

2.2.2. Mechanical propertiesChanges of mechanical properties during the oxidation of

epoxy/amine have been followed by tensile testing on an Instron®

5881 machine. Dumbells samples have been obtained by cutting100 mm films. The effective zone has a dimension of 30 mm lengthand 4 mm width for DGEBA/TTDA and 10 mm length and 2 mmwidth for DGEBA/IDPA and DGEBU/IPDA. The crosshead speed wasat 1 mm/min was constant during the test and a 100 N cell wasused. Tests have been performed in a controlled environment at23 �C on five samples.

2.2.3. Mass lossWeight changes have been monitored during the thermal

oxidation of epoxy resin thanks to the Dual Range XS105 scale fromMettler Toledowith an accuracy of ± 0.01mg. Thosemeasurementshave been duplicated on two samples with an initial mass ofapproximately 100 mg.

2.2.4. Glass transition measurementDifferential scanning calorimetry measurements were made

with a DSC Q1000 (TA Instruments). Samples with mass rangingbetween 3 and 5mg are sealed in aluminumpanswere heated from0 �C to 250 �C at a 10 �C/min ramp under nitrogen flow (50ml/min).Results were interpreted using TA Analysis software. DSC analyseswere done to check the total cure of samples and to measure thevalue of the glass transition temperature of aged samples. Tg valueswere measured during the second heating ramp (i.e. after havingremoved the thermal history of samples). Two samples of the sameresin have been analyzed for each exposure time to control mea-surements reproducibility.

2.2.5. Sub glass mobilityTo put in evidence sub glass transition, i.e. b transition,

dynamical mechanical analysis (DMA) measurements were per-formed between �100 and 40 �C at 1 Hz on films samples. For thispurpose a DMA Q800 from TA instruments was used in tensionmode.

3. Results

3.1. Mechanical properties changes

To put evidence drop of plastic deformation by oxidation,samples were characterized by classical tensile tests up to failure atroom temperature. Fig. 2 shows stress-strain curves for all epoxynetworks during exposure at 110 �C. The stress-strain curves showa classical shape for epoxies [14e16] fromwhich the strain at breakwas determined to monitor plastic deformation.

First of all, for DGEBA/IPDA and DGEBA/TTDA networks (Fig. 2aand b), a decrease of strain at break is associated to a modulus in-crease up to complete embrittlement whereas stress-strainchanges for DGEBU/IDPA are clearly more complex.

First, a modulus increase is observed during thermal oxidationin the case of DGEBA/IPDA and DGEBA/TTDA. This phenomenon, socalled anti-plasticization phenomenon, has already been observedduring the epoxy cured with aromatic amine during oxidation [1]where modulus increase is related to decrease in local mobilityassociated to b transition.

Strain at break changes clearly shows an embrittlement processin the case of DGEBA/IPDA andDGEBA/TTDAwhereas strain at break

Fig. 2. Stress-strain curves of (a) DGEBA/IPDA, (b) DGEBA/TTDA and (c) DGEBU/IPDA after several aging times.

changes for DGEBU/IPDA are surprisingly more complicated. To putin evidence the embrittlement process for each system, strain atbreak values as a function of exposure time are reported in Fig. 3:

- for DGEBA/IPDA, it decreases continuously. The time at whichstrain at break is half is initial value is ca 200 h.

Fig. 3. Changes of mechanical properties: strain at break as a function of exposuretime.

- for DGEBA/TTDA, the initially low strain at break value decreasesat a low rate. The time at which εR ¼ εR0/2 is ca 2200 h.

- last, DGEBU/IPDA displays the most surprising trend: εR in-creases first very strongly and reaches values ca 100%. After asort of plateau, it decreases to a value close to εR0/2 after ca1630 h.

3.2. Mass changes vs embrittlement

In the introduction part, we have mentioned that mass loss isoften used as a tracer for embrittlement. In other words, it is pro-posed that samples become brittle when mass loss exceeds 10%. Tocheck the validity of this assumption, we have followed masschanges during exposure in the same conditions. Mass changes as afunction of exposure time are displayed in Fig. 4. In this latter,characteristic time for embrittlement (time for εR ~ εR0/2) is alsoreported for each networks by an arrow.

According to Fig. 4, DGEBU/IPDA displays an important massloss (more than 20% after 1000 h of exposure at 110 �C). Thisnetwork is by far themost unstable system from themass loss pointof view. Indeed, embrittlement occurs at a mass loss close to 2% and12% for DGEBA/IPDA and DGEBA/TTDA respectively.

As a result, it is obvious there is no direct correlation between anonly one value of mass loss and the embrittlement time for thethree systems under investigation. It means that the three systemsstudied here differ by the consequences on oxidation on the ar-chitecture of the network that will be studied below.

Fig. 4. Mass changes under atmospheric air at 110 �C. Arrows indicate the embrittle-ment time.

3.3. Glass transition changes vs embrittlement

During oxidation, it is expected that epoxy network undergoesmacromolecular changes during oxidation since a chain scissionprocess occurs. Since it is challenging to detect the chain scissionsin networks, measurement of glass transition temperature isparticularly interesting since Tg is related to the concentration inelastically active chains [14,17], which decreases with chain scis-sions and increases with crosslinkings. Tg changes during exposureand characteristic time for embrittlement (arrow) are reported inFig. 5, which shows:

- DGEBA/IPDA displays a significant Tg drop revealing a predom-inant chain scission process.

- DGEBA/TTDA displays a significant Tg increase revealing a pre-dominant crosslinking process.

- DGEBU/IPDA show first a slight Tg drop followed by an increase.These changes can be attributed to chain scission events at lowdegree of conversion followed by a majority crosslinking process.

Interestingly, it appears the common point of both networkswhich undergo crosslinking (DGEBA/TTDA and DGEBU/IPDA) is the

Fig. 5. Tg-Tg0 changes versus time at 110 �C in air (arrows indicate embrittlementtime).

presence of methylenic sequences. A quantitative study and mech-anism proposals of all these aspects can be found elsewhere [13].

3.4. Sub-glassy transition

Since two epoxy networks (DGEBA/IPDA and DGEBA/TTDA)remain in glassy state at room temperature even after oxidationprocess, it seems to us that sub-glass transition (so called b tran-sition) has to be investigated for both systems. Tan d spectra ob-tained by DMA are shown in Fig. 6 for several exposure times.

First, a clear b transition is put in evidence between �60and�20 �C for both epoxy networks. Secondly, it appears that tan damplitude decreases clearly for exposure times increasing. Thesespectra modifications could be associated to oxidation DGEBAsegments. Indeed hydroxypropylether groups in DGEBA can beeasily oxidized since they contain some unstable sites such as CHgroups or CH2 groups in alpha position of heteroatoms [18].

4. Discussion

The main objective of this study is to propose mechanismsresponsible for embrittlement process during oxidation. In terms ofstructure-property relationships, different approaches have beenproposed in the literature to relate macromolecular architectureand failure properties: gap between Tg and temperature for me-chanical testing, molar mass between crosslinks mechanicallyactive (MC) or sub glass transition intensity. To investigate the mostrelevant parameter, in other words relevant whatever the epoxy

Fig. 6. Tan d spectra during exposure at 110 �C for (a) DGEBA/IPDA and (b) DGEBA/TTDA.

networks and the macromolecular modifications (chain scission orcrosslinking) occurring during oxidation, we propose here to checkout these three approaches successively.

4.1. Gap between Tg and temperature for mechanical testing

To check out the first approach, strain at break has been plottedas a function of difference between Tg measured and imposedtemperature for tensile testing, room temperature (RT) in this study(Fig. 7). First, since initial Tg-TRT value for DGEBU/IPDA system isbelow 30 �C and decreases below 20 �C during oxidation up to1000 h time of exposure, the mechanical behavior of this systemtends to a rubbery behavior which explains large deformation (seefor instance the stress-strain curve at 1006 h in Fig. 2c). At highoxidation level, the Tg-TRT increase leads to return to a glassy stateand thus to strain at break values. It is noteworthy that the particularDGEBU/IPDA behavior during oxidation requires low initial Tg-TRTvalue and a chain scission followed by a crosslinking mechanism.

In the case of DGEBA/TTDA network, even if the gap at the initialstate is low (45 �C), this gap is inversely proportional to strain atbreak. In a same time, strain at break drops with the gap valueinitially close to 140 �C for the DGEBA/IPDA. As a result, the gapbetween Tg and RT is not a common feature to explain embrittle-ment for both systems.

4.2. Molar mass between crosslinks

It has been proposed that properties at failure are proportionalto MC

1/2 with MC average molar mass between mechanically activecrosslinks [9]. This relationship has been also experimentally vali-dated for rubbers and is based on the fact that strain at break isrelated to maximum stretching of chains between crosslinks. Thefailure then occurs when chain stretching reaches its limit of chainextensibility defined as the square root of the number of crosslinks[19]. One can notice this approach in the case of glassy networksimplies that large deformation is related to cooperative motions.

To assess MC values in our case, a possible way is to useDiMarzio's approach to estimate crosslink density (n) as following:

Tg ¼ Tgl1� ðKDMFnÞ

With KDM is the DiMarzio's constant ca 2.91 for epoxy/amine

Fig. 7. Strain at break as a function of Tg e TRT during exposure at 110 �C for allnetworks (TRT is the room temperature).

trifunctional networks [20], n is the crosslink density (mol kg�1), Tglis the glass transition of the corresponding « virtual » linear poly-mer and F is the Flex parameter (kg mol�1) related to the molarmass per flexible bond. Validity and physical meaning for the net-works under study have been reviewed in Refs. [20,21].

Molar mass between crosslink (MC) can then be obtained fromcrosslink density (n) by:

MC ¼ 2fn

Where f is the network functionality equal to 3 in our case.In Fig. 8, we have reported strain at break as a function of MC

1/2.Concerning DGEBA/TTDA networks, a good correlation is obtained.However, DGEBA/IPDA follows an opposite trend: strain at breakdecreases with MC increasing. As a result, even if this approachcould be consistent for DGEBA/TTDA network possibly because itslow Tg value promotes cooperative motions, DGEBA/IPDA networkshows a counterexample leading to conclude that MC is not therelevant parameter to monitor embrittlement.

4.3. Sub glass transition intensity

According to previously published results [22] evidencing thatimpact strength is governed by sub glass transition so called btransition, sub glass transition depletion could be associated tostrain at break (and related quantities) decrease [23e26]. To checkout this scenario, Fig. 9 below shows strain at break as a function ofthe intensity (area) of the b transition. It is observed that the plasticdeformation decreases with the area the b transition. This scenariothus appears valid to describe not only the embrittlement ofDGEBA/TTDA but also, and especially, the embrittlement of DGEBA/IPDA. However, the linear characteristic for this relationship be-tween ductility and b transition amplitude needs to be confirmed inthe future.

5. Conclusions

By following physical modifications of three epoxy amine net-works during oxidation, we have investigated the most relevantphysical parameter responsible for embrittlement. Among theseepoxy systems, DGEBA/TTDA has a low initial Tg and undergoesmainly crosslinking whereas DGEBA/IPDA shows a relatively high

Fig. 8. Strain at break as a function of MC1/2.

Fig. 9. Strain at break as a function of b transition amplitude.

Table 1Effect of chain scission on volatile generation in DGEBA/IPDA.

1 or 10 2 or 20 3 or 30 or 300 4 or 40 propability of volatile generation

* * * 1/3* * * 1

* * 1/3

Table 2Effect of chain scission on volatile generation in DGEBA/TTDA.

1 or 10 2 or 20 3 or 30 or 300 4 or 40 5 or 50 propability of volatile generation

* * * 0* * * 1

* * 1* * 1

* * 1* * * 1

Tg and undergoes mainly chain scission. The third chosen system(DGEBU/IPDA) has an initial Tg close to room temperature andshows a chain scission process followed by a crosslink process athigh oxidation level. Firstly, it has been proven that mass changescannot be considered as a good parameter to monitor properties atfailure. Indeed, DGEBA/IPDA becomes brittle for a mass loss close to2%, embrittlement of DGEBA/TTDA occurs at 12%.

To put in evidence the relevant parameter according todifferent approaches proposed in the literature, we have followedTg changes focusing especially on the gap between Tg and tem-perature of tensile testing TRT. It appears that only strain at breakchanges for DGEBU/IPDA can be explained by the Tg-TRT evolu-tion. To investigate DGEBA/TTDA and DGEBA/IPDA embrittle-ment, two physical parameters have been correlated to strain atbreak drop. The first one is the molar mass between crosslink MCcalculated from Tg by DiMarzio's theory and the second one is theintensity of peak associated to b transition. As a result, b transi-tion disappearance during oxidation is the only molecularmodification which can be associated to embrittlement for bothepoxy systems. This finding shows clearly that embrittlement isnot directly associated to chain scission or crosslinking mecha-nism. Knowing that the b transition assigned to hydrox-ypropylether groups local mobility is responsible for ductility,embrittlement appears to be governed by oxidation of thesegroups in the DGEBA segment.

Acknowledgements

ANRT (CIFRE N�2013/0356) is gratefully acknowledged forfinancial support.

Appendix

We are here interested in discussing on the relative weight lossobserved in each kind of epoxy/amine system so as to try to justifythem from a structural point of view.

For that purpose, wewill first try to quantify the occurrence of avolatile compound from chain scissions. In the aim of simplicity, wewill first only consider that only CH2 in a position of nitrogen arereactive (having in mind that the effect of the other CH groups isisopropanol will involve the same kind of effects).

From Table 1, it can thus be estimated that 8 scissions in DGEBA/IPDA are needed to generate 1.66 volatiles. This number may belower if we consider that the tertiary C-H in a position of nitrogenrather yields to scissions of 30 and 300 bonds instead of scission of 3bond (whichwould give a >N� radical, that was not observed, to thebest of our knowledge, in epoxy oxidation).

In DGEBA/TTDA (Table 2), the generation of volatile compoundsis favored for at least two reasons:

- the absence of tertiary CH which favors the yield in scissions ofelastically active chains

- the participation of CH2 in a position of oxygen

Last, in DGEBU/IPDA, the previous reasoning would suggest thatyield of volatile would be the same than in DGEBA/IPDA. However,the possibility of chain scissions occurring on both isopropanolfunctions of the same DGEBU group is another significant source forvolatiles.

References

[1] J.-S. Chen, C.K. Ober, M.D. Poliks, Y. Zhang, U. Wiesner, C. Cohen, Controlleddegradation of epoxy networks: analysis of crosslink density and glass tran-sition temperature changes in thermally reworkable thermosets, Polymer 45(2004) 1939e1950.

[2] S. Terekhina, M. Mille, B. Fayolle, X. Colin, Oxidation induced changes inviscoelastic properties of a thermostable epoxy matrix, Polym. Sci. Ser. A 55(10) (2013) 614e624.

[3] R.J. Morgan, F.-.M. Kong, C.M. Walkup, Structure-property relations ofpolyethertriamine-cured bisphenol-A-diglycidyl ether epoxies, Polymer 25(1984) 375e386.

[4] A.C. Grillet, J. Galy, J.-F. G�erard, J.-P. Pascault, Mechanical and viscoelasticproperties of epoxy networks cured with aromatic diamines, Polymer 32(1991) 1885e1891.

[5] A.E. Mayr, W.D. Cook, G.H. Edward, Yielding behaviour in model epoxythermosets d I. Effect of strain rate and composition, Polymer 39 (1998)3719e3724.

[6] F. Fernandez-Nograro, A. Valea, R. Llano-Ponte, I. Mondragon, Dynamic andmechanical properties of DGEBA/poly(propylene oxide) amine based epoxyresins as a function of stoichiometry, Eur. Polym. J. 32 (1996) 257e266.

[7] F.G. Garcia, B.G. Soares, V.J.R.R. Pita, R. S�anchez, J. Rieumont, Mechanicalproperties of epoxy networks based on DGEBA and aliphatic amines, J. Appl.Polym. Sci. 106 (2007) 2047e2055.

[8] E. Espuche, J. Galy, J.-F. G�erard, J.-P. Pascault, H. Sautereau, Influence ofcrosslink density and chain flexibility on mechanical properties of modelepoxy networks, Macromol. Symp. 93 (1995) 107e115.

[9] J.D. LeMay, F.N. Kelley, Structure and ultimate properties of epoxy resins, In:Epoxy Resins and Composites III (ed. Du�sek, K.), Adv. Polym. Sci. 78 (1986)115e148.

[10] E. Urbaczewski-Espuche, J. Galy, J.-F. Gerard, J.-P. Pascault, H. Sautereau, In-fluence of chain flexibility and crosslink density on mechanical properties ofepoxy/amine networks, Polym. Eng. Sci. 31 (1991) 1572e1580.

[11] L. Heux, J.L. Halary, F. Laupretre, L. Monnerie, Dynamical mechanical and 13Cn.m.r. investigations of molecular motions involved in the b relaxation ofepoxy networks based on DGEBA and aliphatic amines, Polymer 38 (8) (1997)1767e1778.

[12] E. Ernault, E. Richaud, B. Fayolle, Thermal-oxidation of epoxy/amine followedby glass transition temperature changes, Polym. Degrad. Stab. 138 (2017)82e90.

[13] E. Ernault, E. Richaud, B. Fayolle, Thermal oxidation of epoxies: influence ofdiamine hardener, Polym. Degrad. Stab. 134 (2016) 76e86.

[14] O. Sindt, J. Perez, J.F. Gerard, Molecular architecture-mechanical behaviourrelationships in epoxy networks, Polymer 37 (1996) 2989e2997.

[15] F. Lapique, K. Redford, Curing effects on viscosity and mechanical properties ofa commercial epoxy resin adhesive, Int. J. Adhesion Adhesives 22 (4) (2002)337e346.

[16] N. Tian, R. Ning, J. Kong, Self-toughening of epoxy resin through controllingtopology of cross-linked networks, Polymer 99 (2016) 376e385.

[17] N. Rasoldier, X. Colin, J. Verdu, M. Bocquet, L. Olivier, L. Chocinski-Arnault,M.C. Lafarie-Frenot, Model systems for thermo-oxidised epoxy compositematrices, Compos. Part A Appl. Sci. Manuf. 39 (2008) 1522e1529.

[18] H. Matsui, C.A. Schehr, J.J. Valentini, J.N. Weber, Resonance Raman Spectros-copy investigation of the mechanism and kinetics of the degradation of N, N-hexamethylene bishexamide, a nylon 6,6 model compound, Polymer 42(2001) 5625e5632.

[19] L.R.G. Treloar, The Physics of Rubber Elasticity, Oxford University Press, 1975.[20] V. Bellenger, J. Verdu, E. Morel, Effect of structure on glass transition tem-

perature of amine crosslinked epoxies, J. Polym. Sci. Part B Polym. Phys. 25(1987) 1219e1234.

[21] Y.-G. Won, J. Galy, J.-P. Pascault, J. Verdu, Prediction of the glass transitiontemperature of cycloaliphatic amineeepoxy networks, J. Polym. Sci. Part BPolym. Phys. 29 (1991) 981e987.

[22] T. Kamon, H. Furukawa, Curing mechanisms and mechanical properties ofcured epoxy resins, In: Epoxy resins and Composites IV (ed. Du�sek, K.), Adv.Polym. Sci. 80 (1986) 173e202.

[23] M. Pecora, Y. Pannier, M.-C. Lafarie-Frenot, M. Gigliotti, C. Guigon, Effect ofthermo-oxidation on the failure properties of an epoxy resin, Polym. Test. 52(2016) 209e217.

[24] J. Decelle, N. Huet, V. Bellenger, Oxidation induced shrinkage for thermallyaged epoxy networks, Polym. Degrad. Stab. 81 (2003) 239e248.

[25] B.J. Anderson, Thermal stability of high temperature epoxy adhesives bythermogravimetric and adhesive strength measurements, Polym. Degrad.Stab. 96 (2011) 1874e1881.

[26] F. Daghia, F. Zhang, C. Cluzel, P. Ladev�eze, Thermo-mechano-oxidativebehavior at the ply's scale: the effect of oxidation on transverse cracking incarboneepoxy composites, Compos. Struct. 134 (2015) 602e612.